Prime Video uses machine learning to ensure video quality
Detectors for block corruption, audio artifacts, and errors in audio-video synchronization are just three of Prime Video’s quality assurance tools.
At Prime Video, we want to delight customers by continuously expanding our content’s quality and diversity. This includes Amazon Originals, other video-on-demand (VOD) content, and live coverage of sporting events such as NFL Thursday Night Football (TNF), NBA basketball, or French Open tennis.
Because customer obsession is woven into the fabric of everything that we do at Prime Video, we want to ensure that our customers have an industry-leading viewing experience. However, the quality of streaming content can be impacted by defects introduced during recording, encoding, packaging, or transmission of content.
At Prime Video, we built detectors for 18 different types of perceptual content defects, including video freezes, stutters, video tearing, synchronization issues between audio and video, and caption quality problems.
Recently, we authored an article for Amazon Science that provides three examples of block corruption, audio artifacts, and audiovisual-synchronization problems, while explaining the technical approach behind each one. We also discuss how our team uses computer vision and machine learning (CV/ML) models to detect audio, video, or A/V sync quality issues that might be perceivable by customers. Finally, we discuss the classic data availability challenge and how we successfully worked around it.
For more in-depth information, see our How Prime Video uses machine learning to ensure video quality article on the Amazon Science website.